p_s/file_summary_by_instance/private.go

   1 // This file contains the library routines for managing the
   2 // file_summary_by_instance table.
   3 package file_summary_by_instance
   4
   5 import (
   6         "database/sql"
   7         "fmt"
   8         "log"
   9         "regexp"
  10         "sort"
  11         "time"
  12
  13         "github.com/sjmudd/pstop/key_value_cache"
  14         "github.com/sjmudd/pstop/lib"
  15 )
  16
  17 /*
  18 CREATE TABLE `file_summary_by_instance` (
  19   `FILE_NAME` varchar(512) NOT NULL,
  20   `EVENT_NAME` varchar(128) NOT NULL,                           // not collected
  21   `OBJECT_INSTANCE_BEGIN` bigint(20) unsigned NOT NULL,         // not collected
  22   `COUNT_STAR` bigint(20) unsigned NOT NULL,
  23   `SUM_TIMER_WAIT` bigint(20) unsigned NOT NULL,
  24   `MIN_TIMER_WAIT` bigint(20) unsigned NOT NULL,
  25   `AVG_TIMER_WAIT` bigint(20) unsigned NOT NULL,
  26   `MAX_TIMER_WAIT` bigint(20) unsigned NOT NULL,
  27   `COUNT_READ` bigint(20) unsigned NOT NULL,
  28   `SUM_TIMER_READ` bigint(20) unsigned NOT NULL,
  29   `MIN_TIMER_READ` bigint(20) unsigned NOT NULL,
  30   `AVG_TIMER_READ` bigint(20) unsigned NOT NULL,
  31   `MAX_TIMER_READ` bigint(20) unsigned NOT NULL,
  32   `SUM_NUMBER_OF_BYTES_READ` bigint(20) NOT NULL,
  33   `COUNT_WRITE` bigint(20) unsigned NOT NULL,
  34   `SUM_TIMER_WRITE` bigint(20) unsigned NOT NULL,
  35   `MIN_TIMER_WRITE` bigint(20) unsigned NOT NULL,
  36   `AVG_TIMER_WRITE` bigint(20) unsigned NOT NULL,
  37   `MAX_TIMER_WRITE` bigint(20) unsigned NOT NULL,
  38   `SUM_NUMBER_OF_BYTES_WRITE` bigint(20) NOT NULL,
  39   `COUNT_MISC` bigint(20) unsigned NOT NULL,
  40   `SUM_TIMER_MISC` bigint(20) unsigned NOT NULL,
  41   `MIN_TIMER_MISC` bigint(20) unsigned NOT NULL,
  42   `AVG_TIMER_MISC` bigint(20) unsigned NOT NULL,
  43   `MAX_TIMER_MISC` bigint(20) unsigned NOT NULL
  44 ) ENGINE=PERFORMANCE_SCHEMA DEFAULT CHARSET=utf8
  45 1 row in set (0.00 sec)
  46 */
  47
  48 //     foo/../bar --> foo/bar   perl: $new =~ s{[^/]+/\.\./}{/};
  49 //     /./        --> /         perl: $new =~ s{/\./}{};
  50 //     //         --> /         perl: $new =~ s{//}{/};
  51 const (
  52         re_encoded = `@(\d{4})` // FIXME - add me to catch @0024 --> $ for example
  53 )
  54
  55 var (
  56         re_one_or_the_other    *regexp.Regexp = regexp.MustCompile(`/(\.)?/`)
  57         re_slash_dot_dot_slash *regexp.Regexp = regexp.MustCompile(`[^/]+/\.\./`)
  58         re_table_file          *regexp.Regexp = regexp.MustCompile(`/([^/]+)/([^/]+)\.(frm|ibd|MYD|MYI|CSM|CSV|par)$`)
  59         re_temp_table          *regexp.Regexp = regexp.MustCompile(`#sql-[0-9_]+`)
  60         re_part_table          *regexp.Regexp = regexp.MustCompile(`(.+)#P#p(\d+|MAX)`)
  61         re_ibdata              *regexp.Regexp = regexp.MustCompile(`/ibdata\d+$`)
  62         re_redo_log            *regexp.Regexp = regexp.MustCompile(`/ib_logfile\d+$`)
  63         re_binlog              *regexp.Regexp = regexp.MustCompile(`/binlog\.(\d{6}|index)$`)
  64         re_db_opt              *regexp.Regexp = regexp.MustCompile(`/db\.opt$`)
  65         re_slowlog             *regexp.Regexp = regexp.MustCompile(`/slowlog$`)
  66         re_auto_cnf            *regexp.Regexp = regexp.MustCompile(`/auto\.cnf$`)
  67         re_pid_file            *regexp.Regexp = regexp.MustCompile(`/[^/]+\.pid$`)
  68         re_error_msg           *regexp.Regexp = regexp.MustCompile(`/share/[^/]+/errmsg\.sys$`)
  69         re_charset             *regexp.Regexp = regexp.MustCompile(`/share/charsets/Index\.xml$`)
  70         re_dollar              *regexp.Regexp = regexp.MustCompile(`@0024`) // FIXME - add me to catch @0024 --> $ (specific case)
  71
  72         cache key_value_cache.KeyValueCache
  73 )
  74
  75 type table_row struct {
  76         FILE_NAME string
  77
  78         COUNT_STAR  uint64
  79         COUNT_READ  uint64
  80         COUNT_WRITE uint64
  81         COUNT_MISC  uint64
  82
  83         SUM_TIMER_WAIT  uint64
  84         SUM_TIMER_READ  uint64
  85         SUM_TIMER_WRITE uint64
  86         SUM_TIMER_MISC  uint64
  87
  88         SUM_NUMBER_OF_BYTES_READ  uint64
  89         SUM_NUMBER_OF_BYTES_WRITE uint64
  90 }
  91
  92 // represents a table or set of rows
  93 type table_rows []table_row
  94
  95 // Return the name using the FILE_NAME attribute.
  96 func (r *table_row) name() string {
  97         return r.FILE_NAME
  98 }
  99
 100 func (r *table_row) headings() string {
 101         return fmt.Sprintf("%10s %6s|%6s %6s %6s|%8s %8s|%8s %6s %6s %6s|%s",
 102                 "Latency",
 103                 "%",
 104                 "Read",
 105                 "Write",
 106                 "Misc",
 107                 "Rd bytes",
 108                 "Wr bytes",
 109                 "Ops",
 110                 "R Ops",
 111                 "W Ops",
 112                 "M Ops",
 113                 "Table Name")
 114 }
 115
 116 // generate a printable result
 117 func (row *table_row) row_content(totals table_row) string {
 118         var name string = row.name()
 119
 120         // We assume that if COUNT_STAR = 0 then there's no data at all...
 121         // when we have no data we really don't want to show the name either.
 122         if row.COUNT_STAR == 0 && name != "Totals" {
 123                 name = ""
 124         }
 125
 126         return fmt.Sprintf("%10s %6s|%6s %6s %6s|%8s %8s|%8s %6s %6s %6s|%s",
 127                 lib.FormatTime(row.SUM_TIMER_WAIT),
 128                 lib.FormatPct(lib.MyDivide(row.SUM_TIMER_WAIT, totals.SUM_TIMER_WAIT)),
 129                 lib.FormatPct(lib.MyDivide(row.SUM_TIMER_READ, row.SUM_TIMER_WAIT)),
 130                 lib.FormatPct(lib.MyDivide(row.SUM_TIMER_WRITE, row.SUM_TIMER_WAIT)),
 131                 lib.FormatPct(lib.MyDivide(row.SUM_TIMER_MISC, row.SUM_TIMER_WAIT)),
 132                 lib.FormatAmount(row.SUM_NUMBER_OF_BYTES_READ),
 133                 lib.FormatAmount(row.SUM_NUMBER_OF_BYTES_WRITE),
 134                 lib.FormatAmount(row.COUNT_STAR),
 135                 lib.FormatPct(lib.MyDivide(row.COUNT_READ, row.COUNT_STAR)),
 136                 lib.FormatPct(lib.MyDivide(row.COUNT_WRITE, row.COUNT_STAR)),
 137                 lib.FormatPct(lib.MyDivide(row.COUNT_MISC, row.COUNT_STAR)),
 138                 name)
 139 }
 140
 141 func (this *table_row) add(other table_row) {
 142         this.COUNT_STAR += other.COUNT_STAR
 143         this.COUNT_READ += other.COUNT_READ
 144         this.COUNT_WRITE += other.COUNT_WRITE
 145         this.COUNT_MISC += other.COUNT_MISC
 146
 147         this.SUM_TIMER_WAIT += other.SUM_TIMER_WAIT
 148         this.SUM_TIMER_READ += other.SUM_TIMER_READ
 149         this.SUM_TIMER_WRITE += other.SUM_TIMER_WRITE
 150         this.SUM_TIMER_MISC += other.SUM_TIMER_MISC
 151
 152         this.SUM_NUMBER_OF_BYTES_READ += other.SUM_NUMBER_OF_BYTES_READ
 153         this.SUM_NUMBER_OF_BYTES_WRITE += other.SUM_NUMBER_OF_BYTES_WRITE
 154 }
 155
 156 func (this *table_row) subtract(other table_row) {
 157         this.COUNT_STAR -= other.COUNT_STAR
 158         this.COUNT_READ -= other.COUNT_READ
 159         this.COUNT_WRITE -= other.COUNT_WRITE
 160         this.COUNT_MISC -= other.COUNT_MISC
 161
 162         this.SUM_TIMER_WAIT -= other.SUM_TIMER_WAIT
 163         this.SUM_TIMER_READ -= other.SUM_TIMER_READ
 164         this.SUM_TIMER_WRITE -= other.SUM_TIMER_WRITE
 165         this.SUM_TIMER_MISC -= other.SUM_TIMER_MISC
 166
 167         this.SUM_NUMBER_OF_BYTES_READ -= other.SUM_NUMBER_OF_BYTES_READ
 168         this.SUM_NUMBER_OF_BYTES_WRITE -= other.SUM_NUMBER_OF_BYTES_WRITE
 169 }
 170
 171 // return the totals of a slice of rows
 172 func (t table_rows) totals() table_row {
 173         var totals table_row
 174         totals.FILE_NAME = "Totals"
 175
 176         for i := range t {
 177                 totals.add(t[i])
 178         }
 179
 180         return totals
 181 }
 182
 183 // clean up the given path reducing redundant stuff and return the clean path
 184 func cleanup_path(path string) string {
 185
 186         for {
 187                 orig_path := path
 188                 path = re_one_or_the_other.ReplaceAllString(path, "/")
 189                 path = re_slash_dot_dot_slash.ReplaceAllString(path, "/")
 190                 if orig_path == path { // no change so give up
 191                         break
 192                 }
 193         }
 194
 195         return path
 196 }
 197
 198 // From the original FILE_NAME we want to generate a simpler name to use.
 199 // This simpler name may also merge several different filenames into one.
 200 func (t table_row) simple_name(global_variables map[string]string) string {
 201
 202         path := t.FILE_NAME
 203
 204         if cached_result, err := cache.Get(path); err == nil {
 205                 return cached_result
 206         }
 207
 208         // @0024 --> $ (should do this more generically)
 209         path = re_dollar.ReplaceAllLiteralString(path, "$")
 210
 211         // this should probably be ordered from most expected regexp to least
 212         if m1 := re_table_file.FindStringSubmatch(path); m1 != nil {
 213                 // we may match temporary tables so check for them
 214                 if m2 := re_temp_table.FindStringSubmatch(m1[2]); m2 != nil {
 215                         return cache.Put(path, "<temp_table>")
 216                 }
 217
 218                 // we may match partitioned tables so check for them
 219                 if m3 := re_part_table.FindStringSubmatch(m1[2]); m3 != nil {
 220                         return cache.Put(path, m1[1]+"."+m3[1]) // <schema>.<table> (less partition info)
 221                 }
 222
 223                 return cache.Put(path, m1[1]+"."+m1[2]) // <schema>.<table>
 224         }
 225         if re_ibdata.MatchString(path) == true {
 226                 return cache.Put(path, "<ibdata>")
 227         }
 228         if re_redo_log.MatchString(path) == true {
 229                 return cache.Put(path, "<redo_log>")
 230         }
 231         if re_binlog.MatchString(path) == true {
 232                 return cache.Put(path, "<binlog>")
 233         }
 234         if re_db_opt.MatchString(path) == true {
 235                 return cache.Put(path, "<db_opt>")
 236         }
 237         if re_slowlog.MatchString(path) == true {
 238                 return cache.Put(path, "<slow_log>")
 239         }
 240         if re_auto_cnf.MatchString(path) == true {
 241                 return cache.Put(path, "<auto_cnf>")
 242         }
 243         // relay logs are a bit complicated. If a full path then easy to
 244         // identify,but if a relative path we may need to add $datadir,
 245         // but also if as I do we have a ../blah/somewhere/path then we
 246         // need to make it match too.
 247         if len(global_variables["relay_log"]) > 0 {
 248                 relay_log := global_variables["relay_log"]
 249                 if relay_log[0] != '/' { // relative path
 250                         relay_log = cleanup_path(global_variables["datadir"] + relay_log) // datadir always ends in /
 251                 }
 252                 re_relay_log := relay_log + `\.(\d{6}|index)$`
 253                 if regexp.MustCompile(re_relay_log).MatchString(path) == true {
 254                         return cache.Put(path, "<relay_log>")
 255                 }
 256         }
 257         if re_pid_file.MatchString(path) == true {
 258                 return cache.Put(path, "<pid_file>")
 259         }
 260         if re_error_msg.MatchString(path) == true {
 261                 return cache.Put(path, "<errmsg>")
 262         }
 263         if re_charset.MatchString(path) == true {
 264                 return cache.Put(path, "<charset>")
 265         }
 266         // clean up datadir to <datadir>
 267         if len(global_variables["datadir"]) > 0 {
 268                 re_datadir := regexp.MustCompile("^" + global_variables["datadir"])
 269                 path = re_datadir.ReplaceAllLiteralString(path, "<datadir>/")
 270         }
 271
 272         return cache.Put(path, path)
 273 }
 274
 275 // Convert the imported "table" to a merged one with merged data.
 276 // Combine all entries with the same "FILE_NAME" by adding their values.
 277 func merge_by_table_name(orig table_rows, global_variables map[string]string) table_rows {
 278         start := time.Now()
 279         t := make(table_rows, 0, len(orig))
 280
 281         m := make(map[string]table_row)
 282
 283         // iterate over source table
 284         for i := range orig {
 285                 var file_name string
 286                 var new_row table_row
 287                 orig_row := orig[i]
 288
 289                 if orig_row.COUNT_STAR > 0 {
 290                         file_name = orig_row.simple_name(global_variables)
 291
 292                         // check if we have an entry in the map
 293                         if _, found := m[file_name]; found {
 294                                 new_row = m[file_name]
 295                         } else {
 296                                 new_row.FILE_NAME = file_name
 297                         }
 298                         new_row.add(orig_row)
 299                         m[file_name] = new_row // update the map with the new value
 300                 }
 301         }
 302
 303         // add the map contents back into the table
 304         for _, row := range m {
 305                 t = append(t, row)
 306         }
 307
 308         lib.Logger.Println("merge_by_table_name() took:", time.Duration(time.Since(start)).String())
 309         return t
 310 }
 311
 312 // Select the raw data from the database into table_rows
 313 // - filter out empty values
 314 // - merge rows with the same name into a single row
 315 // - change FILE_NAME into a more descriptive value.
 316 func select_rows(dbh *sql.DB) table_rows {
 317         var t table_rows
 318         start := time.Now()
 319
 320         sql := "SELECT FILE_NAME, COUNT_STAR, SUM_TIMER_WAIT, COUNT_READ, SUM_TIMER_READ, SUM_NUMBER_OF_BYTES_READ, COUNT_WRITE, SUM_TIMER_WRITE, SUM_NUMBER_OF_BYTES_WRITE, COUNT_MISC, SUM_TIMER_MISC FROM file_summary_by_instance"
 321
 322         rows, err := dbh.Query(sql)
 323         if err != nil {
 324                 log.Fatal(err)
 325         }
 326         defer rows.Close()
 327
 328         for rows.Next() {
 329                 var r table_row
 330
 331                 if err := rows.Scan(&r.FILE_NAME, &r.COUNT_STAR, &r.SUM_TIMER_WAIT, &r.COUNT_READ, &r.SUM_TIMER_READ, &r.SUM_NUMBER_OF_BYTES_READ, &r.COUNT_WRITE, &r.SUM_TIMER_WRITE, &r.SUM_NUMBER_OF_BYTES_WRITE, &r.COUNT_MISC, &r.SUM_TIMER_MISC); err != nil {
 332                         log.Fatal(err)
 333                 }
 334                 t = append(t, r)
 335         }
 336         if err := rows.Err(); err != nil {
 337                 log.Fatal(err)
 338         }
 339         lib.Logger.Println("select_rows() took:", time.Duration(time.Since(start)).String())
 340
 341         return t
 342 }
 343
 344 // remove the initial values from those rows where there's a match
 345 // - if we find a row we can't match ignore it
 346 func (this *table_rows) subtract(initial table_rows) {
 347         i_by_name := make(map[string]int)
 348
 349         // iterate over rows by name
 350         for i := range initial {
 351                 i_by_name[initial[i].name()] = i
 352         }
 353
 354         for i := range *this {
 355                 if _, ok := i_by_name[(*this)[i].name()]; ok {
 356                         initial_i := i_by_name[(*this)[i].name()]
 357                         (*this)[i].subtract(initial[initial_i])
 358                 }
 359         }
 360 }
 361
 362 func (t table_rows) Len() int      { return len(t) }
 363 func (t table_rows) Swap(i, j int) { t[i], t[j] = t[j], t[i] }
 364 func (t table_rows) Less(i, j int) bool {
 365         return (t[i].SUM_TIMER_WAIT > t[j].SUM_TIMER_WAIT) ||
 366                 ((t[i].SUM_TIMER_WAIT == t[j].SUM_TIMER_WAIT) && (t[i].FILE_NAME < t[j].FILE_NAME))
 367 }
 368
 369 func (t *table_rows) sort() {
 370         sort.Sort(t)
 371 }
 372
 373 // if the data in t2 is "newer", "has more values" than t then it needs refreshing.
 374 // check this by comparing totals.
 375 func (t table_rows) needs_refresh(t2 table_rows) bool {
 376         my_totals := t.totals()
 377         t2_totals := t2.totals()
 378
 379         return my_totals.SUM_TIMER_WAIT > t2_totals.SUM_TIMER_WAIT
 380 }